How do you identify, prioritize, and pay down technical debt?

Technical debt is the gap between how the code is and how it should be, taken on (deliberately or not) to move faster. Some debt is fine; the job is to manage it, not eliminate it. The key is prioritizing the debt that actually hurts.

Prioritize by impact, not annoyance

For each piece of debt, ask:
  - How often does it slow us down or cause bugs? (frequency)
  - How bad is it when it bites? (severity)
  - How risky/expensive is it to fix? (cost)

High frequency + high severity + low cost → fix now
Low frequency + low severity → leave it, document it

A concrete example